How to Measure AI Visibility Across 5 Search Engines
Measuring AI visibility requires tracking five dimensions (mentions, citations, position, sentiment, and source types) across all five major AI engines (ChatGPT, Perplexity, Gemini, Grok, and Claude) at a minimum 48-hour cadence. Single-engine snapshots are unreliable: AI engines disagree on the top recommendation in 50% of queries, and citation counts can swing 48% between identical runs of the same query. A measurement framework that checks fewer than five engines or measures less frequently than weekly will produce data that actively misleads your strategy.
This guide covers the practical mechanics: which engines to track, what to measure, how often to measure it, and the pitfalls that make most teams' data unreliable.
The five engines that matter
As of March 2026, five AI search engines represent the vast majority of AI-driven discovery:
ChatGPT (OpenAI). The largest user base. Uses web search to retrieve current sources and cites them inline. Links to brand sites 18.4% of the time, the highest direct-link rate of any engine.
Perplexity. Built specifically as an AI search engine, not a chatbot with search bolted on. Strong citation behavior with source links displayed prominently. Particularly responsive to Reddit content and recent publications.
Gemini (Google). Google's AI assistant. Integrates with Google's search index, which gives it access to the deepest web corpus. Citation behavior differs from traditional Google search results.
Grok (xAI). Integrated with X (Twitter) data. Cites Reddit content. Links to brand sites 8.5% of the time. Has a distinct personality and response style that affects how brands are presented.
Claude (Anthropic). Conservative citation behavior. Links to brand sites only 3.8% of the time. Cites zero Reddit content. The most restrained of the five engines in terms of external linking.
For a detailed breakdown of each engine's behavior, see The 5 Major AI Search Engines.
Tracking fewer than five engines is a mistake. Here's why.
Why single-engine measurement misleads
AI engines disagree on the top recommendation in 50% of queries. When we ran identical queries across all five engines, pairwise overlap between any two engines ranged from 58% to 75%. A brand that appears in ChatGPT's answer has a roughly 60% to 75% chance of also appearing in Perplexity's answer for the same query. That's a 25% to 40% chance it doesn't.
This means measuring only ChatGPT (or only Perplexity, or only any single engine) gives you at most three-quarters of the picture. The other quarter is either missing brands you're visible on, or missing engines where you're invisible.
The practical impact: a team that tracks only ChatGPT might conclude their AI visibility is strong because they show up in 8 out of 10 tracked queries. But they're invisible on Gemini for 4 of those 10 queries. A prospect using Gemini never sees them. Single-engine data creates blind spots that directly cost you discovery.
Track all five. Always.
What to measure: the five dimensions
AI visibility isn't a single number. It breaks down into five measurable dimensions.
1. Mentions
The most basic signal. Does the AI engine name your brand in its response? Count mentions per query, per engine. Track whether mentions are in a positive, negative, or neutral context.
A mention is not the same as a recommendation. Being listed seventh in a response that names ten competitors is a mention. Being named as the "top option" is a recommendation. The distinction matters for understanding your actual position.
In our research, Netlify accumulated 14 mentions across queries without a single position-1 placement. High mention counts with low position indicate visibility without authority. You're in the conversation but not leading it.
2. Citations
Does the engine link to your content as a source? This is a stronger signal than mentions because it indicates the engine trusts your content enough to reference it. But citation behavior varies dramatically by engine.
| Engine | Links to brand sites |
|---|---|
| ChatGPT | 18.4% of citations |
| Grok | 8.5% of citations |
| Claude | 3.8% of citations |
And across all engines, 92.5% of citations go to third-party sources. Your brand might get mentioned, but the link goes to a G2 review, a TechCrunch article, or a comparison blog post someone else wrote.
This means tracking citations requires tracking both direct citations (links to your content) and indirect citations (mentions sourced from third-party content about you). Both contribute to visibility, but they require different strategies to improve.
3. Position
Where in the response does your brand appear? The first brand named in an AI engine's answer carries the strongest implicit endorsement. Position 5 in a list of 8 is functionally invisible to most users who stop reading after the first few recommendations.
Position tracking is harder than ranking tracking in SEO because AI responses don't have fixed slots. The response is prose, not a list. But you can approximate position by measuring: are you in the first paragraph? First three brands mentioned? First recommendation given?
4. Sentiment
How does the engine characterize your brand? "X is a strong option for startups" is different from "X is an alternative, though less established." Both are mentions. One builds credibility. The other implies weakness.
Sentiment analysis across AI engines requires reading the full context of each mention, not just detecting brand name presence. Automated sentiment scoring works for volume, but manual review catches the nuances that matter.
5. Source types
What kind of sources is the engine citing when it mentions your brand? Your own blog posts? Third-party reviews? Forum discussions? News articles?
Source type analysis tells you where your authority is coming from. If all your citations trace back to a single G2 review page, your visibility is fragile: if that page changes, your citations change with it. Diverse source types indicate more stable visibility.
Grok cites Reddit content. Claude cites zero Reddit. Perplexity is responsive to recent publications. Each engine has source preferences that affect your strategy.
The snapshot problem: why frequency matters
Here is the single most important measurement principle: a single snapshot of AI visibility is unreliable.
In our research, citation counts swung 48% between identical runs of the same query on the same engine. Run a query on Monday, get cited three times. Run the exact same query on Wednesday, get cited once. Or five times. The variance is real and significant.
This happens because AI engines are nondeterministic. Their retrieval systems don't return the exact same set of sources every time. The models introduce randomness in generation. Knowledge base updates add or remove sources from the retrieval pool. For a deeper technical explanation, see why AI search engine citations are nondeterministic.
The practical implication: you need multiple measurements over time to establish a reliable signal. A single check that shows your brand wasn't mentioned doesn't mean you have zero visibility. It means you had zero visibility in that specific run. Check again tomorrow and you might appear.
The FogTrail AEO platform's 48-hour intelligence cycles exist precisely because of this variance. By rechecking queries every 48 hours across all five engines, the system builds a rolling picture of your visibility that smooths out the noise of individual runs. A trend line over two weeks is meaningful. A single data point is not.
How often to measure
The minimum useful cadence is weekly. Checking monthly gives you data that's already stale, because engine knowledge bases refresh faster than your measurement cycle.
The optimal cadence is every 48 hours, aligned with the approximate refresh rate of engine knowledge bases. This lets you catch visibility changes before they compound and respond within the same refresh window.
Daily measurement is technically possible but creates noise. The nondeterministic variance between runs means daily data points jump around. Every-other-day smooths this while still being responsive.
Building a measurement framework
Here's a practical framework for measuring AI visibility across five engines.
Step 1: Define your query set
Start with 20 to 50 queries that represent how your target audience discovers solutions in your category. Include:
- Category queries ("best [category] tools")
- Problem queries ("how to solve [problem your product addresses]")
- Comparison queries ("[your brand] vs [competitor]")
- Use case queries ("[your product type] for [specific use case]")
Avoid vanity queries (your own brand name). You want to measure discoverability, not name recognition.
Step 2: Run queries across all five engines
For each query, record the full response from each engine. Extract: mentions (which brands appeared), citations (which sources were linked), position (order of brand mentions), and sentiment (how each brand was characterized).
This is time-intensive if done manually. For 50 queries across 5 engines, that's 250 individual checks per measurement cycle. At a 48-hour cadence, that's 875 checks per week. Automation is not optional at scale.
Step 3: Calculate your metrics
Mention rate: Percentage of query/engine combinations where your brand appears. Example: mentioned in 120 of 250 checks = 48% mention rate.
Citation rate: Percentage of mentions that include a direct citation (link) to your content. Low citation rate with high mention rate means engines know about you but don't trust your content enough to cite it directly.
Position-1 rate: Percentage of mentions where your brand is the first named. This is your authority signal.
Engine coverage: How many of the five engines mention you for each query. Full coverage (5/5) means you're visible everywhere. Partial coverage (2/5 or 3/5) means engine-specific gaps.
Competitive share: For each query, how many total brand mentions exist, and what percentage are yours. If an engine mentions six brands and you're one of them, your share is 16.7%. Track this against specific competitors.
For a comprehensive breakdown of AEO metrics, see AEO metrics and KPIs.
Step 4: Trend over time
Plot these metrics across measurement cycles. The absolute numbers matter less than the direction. Are your mention rates climbing? Is your citation rate improving? Are you gaining ground on competitors?
The trend is the signal. Individual data points are noise.
PostHog's trajectory in our research illustrates this: citations grew from 2 to 3 to 5 across three measurement waves. That's a clear upward trend. If you only measured once, you'd see 2, or 3, or 5, with no context for which.
Step 5: Segment by engine
Don't just look at aggregate numbers. Break down every metric by engine. You might be strong on ChatGPT and invisible on Claude. You might be cited by Perplexity but never by Grok.
Engine-specific gaps require engine-specific strategies. Claude's conservative citation behavior means different content optimization than Perplexity's aggressive sourcing. Multi-engine AEO covers the engine-by-engine approach.
Common measurement mistakes
Measuring your own brand name. If you search "What is [your brand]?" and you appear in the answer, that tells you nothing useful. Of course the engine knows your name. Measure category queries where you compete for visibility.
Treating a single run as truth. One check that shows zero mentions does not mean zero visibility. Run it again. And again. Build a sample before drawing conclusions.
Ignoring third-party citations. If the engine cites a G2 review that mentions you, that's still AI visibility. Don't only track direct citations to your own domain.
Measuring too infrequently. Monthly checks miss the 48-hour refresh cycle of engine knowledge bases. You could publish content, have it cited for two weeks, and lose the citation before your next measurement. Weekly minimum, 48-hour cadence ideal.
Not controlling for nondeterminism. Some teams run a query once, see they're cited, and celebrate. Then they run it again and the citation is gone. The 48% variance between identical runs means you need multiple samples to establish a reliable baseline. See nondeterministic citations for the full data.
From measurement to action
Measurement is necessary but insufficient. The purpose of measuring AI visibility is to identify what needs to change, then change it. The metrics tell you where you're strong and where you're weak. The action plan tells you what to do about the weak spots.
If your mention rate is low, you have a content existence problem. There isn't enough content about your brand for engines to draw from.
If your mention rate is high but your citation rate is low, you have a content quality problem. Engines know about you but don't trust your content enough to cite it directly.
If your position-1 rate is low, you have an authority problem. Engines cite competitors first.
If your engine coverage is spotty, you have a distribution problem. Some engines can find you, others can't.
Each of these diagnoses maps to a different set of actions. Content creation, structural optimization, third-party coverage, engine-specific targeting. What to do after AEO monitoring provides the next-step playbook for each scenario.
The FogTrail AEO platform's measurement backbone
The FogTrail AEO platform automates this entire measurement framework. Every 48 hours, the platform rechecks your full query set across all five engines, calculates the metrics described above, identifies changes from the previous cycle, and generates intelligence briefings that explain what changed and why.
The briefings include competitive narrative analysis (what stories your competitors are winning), specific action proposals (what content to create), and verification tracking (whether previous content actually improved citations). It's measurement wired directly to execution.
As of March 2026, the FogTrail AEO platform is $499/mo. 100 queries across 5 engines. 48-hour cycles. The measurement is the foundation, but the execution on top of it is what actually changes your AI visibility.
Frequently Asked Questions
How often should I measure AI visibility across engines?
The minimum useful cadence is weekly. The optimal cadence is every 48 hours, aligned with the approximate refresh rate of engine knowledge bases. Daily measurement creates noise due to nondeterministic variance between runs. Every-other-day smooths this while still being responsive enough to catch changes before they compound.
Can I measure AI visibility with just one engine?
No. AI engines disagree on the top recommendation in 50% of queries, and pairwise overlap between any two engines ranges from 58% to 75%. Measuring only one engine gives you at most three-quarters of the picture. A brand visible on ChatGPT may be invisible on Gemini for the same query. Track all five engines to avoid blind spots.
What is the most important AI visibility metric to track?
Position-1 rate, the percentage of mentions where your brand is the first named, is the strongest authority signal. High mention counts with low position indicate visibility without authority. But no single metric tells the full story. Combine mention rate, citation rate, position, sentiment, and engine coverage for a complete picture.
How reliable is a single AI visibility snapshot?
Not reliable at all. Citation counts can swing 48% between identical runs of the same query on the same engine. A single check that shows your brand was not mentioned does not mean you have zero visibility. You need multiple measurements over time to establish a reliable signal. A trend line over two weeks is meaningful. A single data point is not.
What is the difference between a mention and a citation in AI search?
A mention means the AI engine names your brand in its response. A citation means the engine links to your content as a source. Citations are a stronger signal because they indicate the engine trusts your content enough to reference it directly. Across all engines, 92.5% of citations go to third-party sources, so tracking both direct citations (links to your content) and indirect citations (mentions sourced from third-party content about you) matters.